A comparison of manual and automated methods of measuring conjunctival vessel widths from photographic and digital images

Owen, Christopher G., Ellis, Timothy J. and Woodward, E. Geoffrey (2004) A comparison of manual and automated methods of measuring conjunctival vessel widths from photographic and digital images. Ophthalmic And Physiological Optics, 24(2), pp. 74-81. ISSN (print) 0275-5408

Full text not available from this archive.

Abstract

We investigated the application of a fully automated computer algorithm for identifying vessels of the conjunctiva from their scleral surround, and compared measures of vessel width with established methods. Vessel widths at 101 locations (ranging from 20 to 140 microm), from 12 patients, were measured from film and digital images, using a variety of methods, and compared. Widths were measured manually, by semi-automated methods using grey level (densitometric) profiles taken from digital images, and by automated techniques set at different operating levels. Good intra-session repeatibility was obtained using the automated method with an operating sigma value of 3 pixels (16 microm) (mean difference 0.5 microm, 95% CI -8.5 to 9.4 microm) and manual calliper measurements from digitally created photographic slides (mean difference 0.4 microm, -9.3 to 10.1 microm). For comparison with other measures of width, the latter was used as the gold standard. Widths measured from film were slightly larger than those measured directly from digital images, although this effect was small (5 microm) for most vessels. Overall widths measured using the automated method, with a sigma value of 3 pixels, agreed best with the gold standard (inter-method repeatibility; mean difference 1.4 microm, -32.5 to 35.2 microm) although the automated method overestimated small widths (<40 microm) and underestimated larger vessel widths (>40 microm). Automated detection of vessels of the conjunctiva from digital images avoids manual and operator involved measures which are time consuming, and which preclude large patient studies. The resulting data may help in monitoring the vascular response of the conjunctiva to surgical or pharmacological intervention, and in describing vascular changes in response to ocular or systemic disease. The application of this algorithm to the study of retinal vessels is yet to be realised.

Item Type: Article
Uncontrolled Keywords: automated, conjunctiva, measurement, vessel width, retinal microvascular abnormalities, atherosclerosis risk, fundus photographs, blood-vessels, circulation, communities, glaucoma
Research Area: Computer science and informatics
Faculty, School or Research Centre: Faculty of Computing, Information Systems and Mathematics (until 2011)
Faculty of Computing, Information Systems and Mathematics (until 2011) > Digital Imaging Research Centre (DIRC)
Related URLs:
Depositing User: Automatic Import Agent
Date Deposited: 15 Jan 2010 10:37
Last Modified: 14 Oct 2010 09:16
URI: http://eprints.kingston.ac.uk/id/eprint/8411

Actions (Repository Editors)

Item Control Page Item Control Page